Hi,

One thing that might be going on is that your CMFB circuit might not have enough input dynamic range to accommodate your large input signal swing. As a consequence the CM level will change since its gain will drop. The valid CM level is the dc operating point one. However you might want to check the CMFB dynamic range to make sure it tolerates the desired input signal amplitude.

Hope this helps
Tosei

Hi CherryQ,

I could imagine that the output you get during TRAN analysis is not pure sinusoidal, which means that its mean value is NOT zero and, therefore, adds to the DC level.
Lutz

Hi CherryQ,

I just meant that you ought to ensure that the common-mode sensing and feedback circuitry is linear enough that large differential
signals at the output of your amplifier do not lead the CMFB going out of its linear range, or cause some differential mode
signal from leaking into the common mode path.

To put it on a concrete footing, if you have outputs Vp and Vn, then your common-mode
sensing circuitry should always detect Vcm = 0.5*(Vp + Vn).

Due to nonidealities or poor choice of architecture, you might pass the signal Vcm = 0.5*(Vp+Vn) + Admcm(Vp - Vn) into
your common-mode path. However, your problem could be more due to the common-mode sensing circuitry having a limited
range of input swing (smaller than swing actually seen in amp).

Vivek